AgenticFinLab; The Hong Kong University of Science and Technology (Guangzhou); Yantai Research Institute of Harbin Engineering University, 2026

PortBench

Multi-asset portfolio management across six asset classes. It pairs a static QA set with a five-stage decision pipeline whose allocation stage is scored on how well it captures cross-asset correlation, under stress windows and three risk profiles.

Best reported

0.384

GLM-5.1Zhipu AI

CEPS (0-1)

Reported by the authors, 9 Oct 2026

About this benchmark

Size
6,269 QA items across 7 templates (4 complexity levels); 5-stage pipeline; 183 market series across 6 asset classes, 2015-2025
Public split
Code on GitHub; dataset as a Hugging Face collection; raw experimental results as a Hugging Face dataset
Grading
QA: mean score per template. Pipeline: CEPS, the mean of five stage scores normalised to [0,1] minus a 0.1 penalty for drops between stages, plus Sharpe, max drawdown, volatility and return
Licence
Apache-2.0 (code); MIT (PortBench-RawData results dataset on HF)
Submission
none
Products allowed
Not stated
Latest version
arXiv v6 (2026-10-09)
What it does not cover
Live execution costs and tool or agent products; the 12-day real-time run is only a preliminary demo

Results

10 models
Leaderboard

Each score shows who reported it: the benchmark's authors (quoted from their paper), the benchmark's official board, or Moneda (run with the benchmark's own grader and rules). We do not copy other aggregators.

PortBench arXiv v6 (2026-10-09)

CEPS (0-1), higher is better. 10 models.

Updated 9 Oct 2026

  1. GLM-5.10.384
  2. Qwen3.6-Plus0.300
  3. Qwen3.7-Max0.292
  4. Kimi K2.60.274
  5. Qwen3.6-35B-A3B0.272
  6. DS-V4-Pro0.266
  7. DB-2.0-Pro0.263
  8. HY3-Preview0.258
  9. DB-2.0-Lite0.244
  10. DS-V4-Flash0.226

PortBench arXiv v6 (2026-10-09)

CEPS (0-1), higher is better. 10 models.

Updated 9 Oct 2026

  1. 1GLM-5.10.384
  2. 2Qwen3.6-Plus0.300
  3. 3Qwen3.7-Max0.292
  4. 4Kimi K2.60.274
  5. 5Qwen3.6-35B-A3B0.272
  6. 6DS-V4-Pro0.266
  7. 7DB-2.0-Pro0.263
  8. 8HY3-Preview0.258
  9. 9DB-2.0-Lite0.244
  10. 10DS-V4-Flash0.226
PortBench leaderboard (CEPS (0-1), higher is better)
1GLM-5.1Zhipu AI0.384pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
2Qwen3.6-PlusQwen0.300pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
3Qwen3.7-MaxQwen0.292pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
4Kimi K2.6Moonshot AI0.274pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
5Qwen3.6-35B-A3BQwen0.272pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
6DS-V4-ProDeepSeek0.266pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
7DB-2.0-ProNot stated0.263pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
8HY3-PreviewTencent0.258pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
9DB-2.0-LiteNot stated0.244pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026
10DS-V4-FlashDeepSeek0.226pipeline, balanced risk profilePaper Table 2
Reported by the authors, 9 Oct 2026

PortBench arXiv v6 (2026-10-09)

mean QA score (0-1), higher is better. 10 models.

Updated 9 Oct 2026

  1. DS-V4-Flash0.819
  2. Qwen3.7-Max0.819
  3. DS-V4-Pro0.818
  4. DB-2.0-Lite0.804
  5. DB-2.0-Pro0.787
  6. Qwen3.6-Plus0.783
  7. GLM-5.10.757
  8. Qwen3.6-35B-A3B0.684
  9. HY3-Preview0.624
  10. Kimi K2.60.511

PortBench arXiv v6 (2026-10-09)

mean QA score (0-1), higher is better. 10 models.

Updated 9 Oct 2026

  1. 1DS-V4-Flash0.819
  2. 1Qwen3.7-Max0.819
  3. 3DS-V4-Pro0.818
  4. 4DB-2.0-Lite0.804
  5. 5DB-2.0-Pro0.787
  6. 6Qwen3.6-Plus0.783
  7. 7GLM-5.10.757
  8. 8Qwen3.6-35B-A3B0.684
  9. 9HY3-Preview0.624
  10. 10Kimi K2.60.511
PortBench leaderboard (mean QA score (0-1), higher is better)
1DS-V4-FlashDeepSeek0.819static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
1Qwen3.7-MaxQwen0.819static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
3DS-V4-ProDeepSeek0.818static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
4DB-2.0-LiteNot stated0.804static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
5DB-2.0-ProNot stated0.787static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
6Qwen3.6-PlusQwen0.783static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
7GLM-5.1Zhipu AI0.757static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
8Qwen3.6-35B-A3BQwen0.684static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
9HY3-PreviewTencent0.624static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026
10Kimi K2.6Moonshot AI0.511static QA, mean over T1-T7 (secondary)Paper Table 1
Reported by the authors, 9 Oct 2026

Last checked 1 Nov 2026. Something out of date or wrong? Send a correction.